Linguistic analysis for emotion recognition: a case of Chinese speakers

Acoustic features, in the examination of emotions occurring in pronouncing English and Chinese Mandarin words, are investigated in this study, then different emotion recognition experiments are presented. To this end, the sound recordings for 91 speakers were analyzed. In the test experiment, a linguistic data set was used to examine which acoustic features are most important for the emotional representation in signal acquisition, segmentation, construction, and encoding. In doing so, words, syllables, phonemes (which contain vowels and consonants), stress and frequency tones were taken into consideration. The types of emotions considered in the experiment included neutral, happy, and sad. Time duration differences, F0 frequency, and dB intensity levels variables were used in conjunction with unsupervised and supervised machine learning approaches for emotion recognition.